Computer worms operating in tandem with free AI chatbots like ChatGPT are a 'new class of cyberthreat', security experts have warned. Worms are a type of malware that self-replicate and spread across computer networks without requiring human interaction, unlike viruses, which need unsuspecting users to open an infected file. They can copy themselves rapidly, clogging system memory and network bandwidth to overload or crash systems. This makes anti-virus software crucial for protection.
One notable example is the WannaCry worm, responsible for a global ransomware attack in 2017. It targeted computers running Microsoft Windows by encrypting data and demanding ransom payments in bitcoin. Now, scientists have demonstrated for the first time that publicly accessible large language AI models, similar to ChatGPT, can be used to power worms that adapt their strategy as they spread from one device to the next.
AI-Driven Worms: A New Threat
'The worm parasitically uses compromised machines to run open-access large language models (LLMs) to sustain its reasoning, or extend its reach for further attacks,' researchers warn in a yet-to-be-peer-reviewed study posted on arXiv. 'Deployed on a network of machines spanning Linux, Windows, and IoT devices, the worm propagated by exploiting common, real-world corporate network vulnerabilities,' they wrote.
In the latest study, conducted at a secure digital lab walled off from the outside world, scientists simulated the capabilities of an AI-driven worm on dozens of interconnected devices, including laptops, printers, and cameras. 'It was imperative for us to understand this threat in a controlled, academic setting before bad actors figured it out for themselves,' said Nicolas Papernot, one of the authors from the University of Toronto. 'The reason we are doing this research is to ensure the security of the digital ecosystem we all rely on – to keep people safe. This finding catapults us into a new era of cybersecurity,' Dr Papernot added.
How AI Worms Operate
A growing body of studies shows that AI models like OpenAI's ChatGPT and Google's Gemini can be tricked into losing their safety guardrails and manipulated to share harmful information. Researchers showed that publicly accessible AI models could be used to engineer far more sophisticated threats, including ones that can scope out their targets and tailor attacks to take over a machine, before cloning itself onto the next one. 'Our results demonstrate that self-sustaining AI-driven cyber-threats are no longer theoretical. We must prepare for autonomous generative adversaries,' scientists wrote.
Worms made using AI can adapt and gather information as they move deeper into a network, with every breach revealing passwords and weak points that can unlock another machine. 'Hackers have typically had to prioritise the most high-value targets because time and computing resources were limited. But now, once a worm is launched, the cost would drop to nearly zero,' Dr Papernot said. 'Every device connected to the internet – laptops, cameras, smart thermostats and everything else – becomes a potential target, if not for the data it holds, then as a foothold to attack more valuable targets.'
Urgent Need for Action
Such AI-powered worms could gain internet access and scan and exploit warning notices about newly discovered vulnerabilities, outpacing the software patches meant to stop them. 'In an interconnected world, no system is immune to this threat. Sharing these findings is the first step in galvanising researchers, industry leaders and policymakers to take action – and quickly,' the cybersecurity expert said. 'We can no longer afford to hit 'ignore' on software updates. Every door you close is one less way in, so it's worth taking a few minutes to reboot,' Dr Papernot warned.
